Competitive Binding of Bilirubin and Drugsto Human Serum Albumin Studied by Enzymatic Oxidation

R. BRODERSEN

From the Institute of Medical Biochemistry, University of Aarhus,8000 Aarhus C, Denmark

A B S T R A C T The mechanism of drug-induced dis-placement of bilirubin from the blood into tissues wasstudied. A model of simple, competitive binding of bili-rubin and drug to one site on serum albumin was es-tablished. Variations of the free bilirubin concentrationafter addition of drugs were studied in vitro by mea-suring velocities of oxidation with hydrogen peroxideand horseradish peroxidase. In all cases, the resultswere in agreement with the model. The competitive ef-fects of 20 drugs were measured and expressed quanti-tatively as binding constants to the bilirubin site on hu-man serum albumin. Several drugs caused changes ofthe bilirubin-albumin light absorption spectrum, indi-cating simultaneous binding of both ligands, without aneffect on the free bilirubin concentration. Noncompetitivesite-to-site effects on bilirubin binding could not bedemonstrated.

An equation is proposed for calculation of the maxi-mal displacing effect of a drug from knowledge of itsplasma concentration, the above-determined binding con-stant, and the degree of protein binding of the drug.

Comparison of these results with previous observationsof bilirubin displacement in newborn humans and inexperimental animals indicates a general agreementwith a simple competitive mechanism of binding ofbilirubin and drug to one site on the albumin molecule.Binding of drugs to other, noncompetitive sites iscommon.

INTRODUCTIONA number of drugs, if given to newborn children withunconjugated hyperbilirubinemia, may displace bilirubinfrom its binding to plasma albumin, with the possibleeffect that the pigment enters the central nervous system,causing kernicterus or various late manifestations ofbrain damage. Sulfisoxazole is known to exert such aneffect clinically (1, 2). This and other sulfonamidesmay precipitate kernicterus in hyperbilirubinemic ani-

Received for publication 21 March 1974 and in revisedform 29 June 1974.

mals (3, 4) and show a bilirubin-displacing effect invivo (3, 5) and in vitro (6, 7). Salicylate, given to hu-mans (8) and to Gunn rats (9, 10), causes a drop ofserum bilirubin concentration, presumably due to dis-placement of the pigment to extravascular receptors.Large doses of fatty acids, infused to a child, resulted ina marked increase of the concentration of unbound bili-rubin (11).

Some investigators have doubted the danger of givingsulfonamides to icteric newborns (12, 13), and a com-petitive effect of salicylate has been questioned by Thalerand Schmid (14). The latter authors, invoking a dif-ferent principle, warn against the use of acetylsalicylicacid in hyperbilirubinemic patients because this drughas been shown to acetylate serum albumin (15), im-pairing its affinity for nonpolar anions. This uncertaintycalls for investigation of the mechanism involved in dis-placement of bilirubin with drugs. In the present paper,the problem is elucidated by kinetic studies of oxidationof free bilirubin with hydrogen peroxide and peroxidase.A method for quantitative assessment of displacing ef-fects is proposed.

As a basis for calculation of the effect of a competingdrug, a model is defined by the following five assump-tions: (a) Free, monomer bilirubin anion is present insolutions containing bilirubin and a molar excess of hu-man serum albumin at neutral pH. The free bilirubinconcentration is very low and cannot be determined byconventional means. Relative values of the free bilirubinconcentration can be measured by the rate of oxidationwith hydrogen peroxide and peroxidase. (b) One mole-cule of bilirubin is bound reversibly to one high-affinitysite on serum albumin. (c) The free bilirubin concen-tration remains small, compared to the total. (d) Onemolecule of the drug is bound competitively to the bili-rubin-binding site. (e) Additional molecules of thecompetitor may be bound to other sites on albumin,without interfering with binding of bilirubin.

It will be shown that this model is sufficient to de-scribe the experimental findings, and the validity of theassumptions will be discussed.

The Journal of Clinical Investigation Volume 54 December 1974-1353-1364 1353

METHODSOxidation of bilirubin with hydrogen peroxide and peroxi-dase follows Michaelis-Menten kinetics (16). Km for bili-rubin at pH 8.2 is 8 X 10' M and less than 2 X 10' Mfor hydrogen peroxide. It is not possible to determine theseconstants at pH 7.4, due to the low solubility of bilirubinin neutral solutions, but they are unlikely to be severalorders of magnitude different from the values at pH 8.2.This means that the rate of oxidation is proportional tothe bilirubin concentration in the nanomolar range and isindependent of the hydrogen peroxide concentration whenthis is of the order of 100 ,uM. A convenient principle ishereby available for relative determinations of free bili-rubin concentrations in equilibrium mixtures. The processesinvolved are

BP B + P,LX

where B, P, and X, are bilirubin, albumin, and oxidationproduct, respectively. The rate of the total process, con-version of the bilirubin-albumin complex to free albuminand the oxidation product, is proportional to the equilibriumconcentration of free bilirubin if the following two condi-tions are fulfilled: (a) the free bilirubin only is oxidized,while bound bilirubin is protected, and (b) the velocityof the dissociation of bilirubin from albumin is high com-pared to the oxidation, so that the latter process is ratedetermining. It has previously been found by Jacobsen (17)that the rate of oxidation varies with the ratio of bilirubinto albumin in accordance with these assumptions, andfurther evidence for their validity is reported in the presentpaper (Figs. 2 and 3).

Addition of a displacing drug to the equilibrium mix-ture causes an increase of the free bilirubin concentrationand an increased oxidation rate. Determination of the ini-tial oxidation rate as a function of drug concentration mayserve as a basis for assessment of the displacing effect ofthe drug. The method is based on the following quantita-tive principles.

P, B, and D denote the total concentrations of albumin,bilirubin, and drug, and [BP] and [DP] the concentrationsof albumin with bilirubin and drug, respectively, bound tothe high affinity site for bilirubin. The concentration of al-bumin without a ligand on this site is p, and the freeconcentrations of bilirubin and drug are b and d. An indexo is used for concentrations in the absence of drug. KBand KD are the binding constants for bilirubin and drug tothe high-affinity bilirubin-site on albumin. We then have

EBP]=

[BP]o = KB, [DP] - KD,pb pobo pd

P = p + [BP] + [DP] = po + [BP]J0and

B = b + [BP] = bo + [BP]o.

For b<<B (see Discussion) we obtain the approxima-tions B n [BP] " [BP]., p. - P-B, [DP] ~ P-B-p, andhence

b = PO=KDd+ 1. (1)bo p

Experimental determinations of KD were based on thisequation. A reaction mixture was prepared, containing bili-rubin and human serum albumin in a molar ratio of 1:2.The velocity of oxidation with hydrogen peroxide andperoxidase was determined with and without added drug.The ratio of these velocities was equated with the ratio offree bilirubin concentrations, b/b,.

The free drug concentration in Eq. 1 is generally notknown. If enough of the drug is added to provide a con-siderable excess over the amount bound to albumin, wecan use the total concentration as an approximation of thefree and obtain

- = KDD+ 1,bo (2)

which is valid when the drug concentration is large, com-pared to that of albumin (see Discussion).

Determination of the initial rate of the oxidation processfrom the slope of the progress curve as soon as possibleafter addition of the enzyme was found to be less re-producible than determination of the time taken for a cer-tain fraction of the total change of optical density. In ex-periments with varying enzyme concentration and with sev-eral drugs, it was found that the ratio of such time periodswith and without added drug was independent of whetherthe fraction 0.1, 0.15, or 0.2 of the total change was used.These ratios are consequently equal to the inverse ratioof initial velocities. Thus

tO.2(0) _ to.lS(o) _ to.l(O) __ v _ b

to.2 to.ls to.l vo bo(3)

when v and v. are the initial velocities with and withoutdrug added. The best reproducibility was obtained by useof to.2.

With increasing concentrations of a competing drug, thevalues of to.2 decrease and the horseradish peroxidase con-centration, [HRP], was lowered to obtain suitable rates.Results from determinations with and without drug added(the latter with index o) were entered in the expression

tO.2(0) [HRP]O - KDD+ 1,

4O.2 [HRP] (4)

which is derived from Eqs. 2 and 3, if the velocity is pro-portional to the enzyme concentration (Fig. 3).

The left side of Eq. 4 was plotted as the ordinates,against the drug concentration D as the abscissa. A linewas fitted with the intercept 1 on the ordinate axis. Theslope of the line is KD, the binding constant of the drug tothe bilirubin site.

Materials. Two batches of human serum albumin wereobtained from AB Kabi, Stockholm, and one (salt: precipi-tated) from Statens Seruminstitut, Copenhagen. The con-

tent of fatty acids was determined according to Dole (18)and was 1.5, 0.9, and 0.5 mol/mol protein, respectively.Oxidation velocities obtained with these albumins were

identical, in agreement with previous observations (19-21)that less than 3 mol fatty acid/mol albumin does not in-fluence the binding of bilirubin. Defatting according toChen (2?) or desalting on a mixed ion exchange bedlikewise did not change the oxidation rates. The albuminpreparations contained a certain amount of dimer and

1354 R. Brodersen

oligomers, of the order of 7%. A few experiments wereconducted with monomer albumin, obtained by Sephadexgel chromatography, with identical results. It has beenfound previously (23) that binding of bilirubin is not in-fluenced by dimerization of the albumin. The derivation ofEq. 1 is also not changed by dimerization of the protein.

Bilirubin was from Sigma Chemical Co., Inc. (St. Louis,Mo.), Eu'% = 1.02 X 103 in chloroform, and was used assuch. Horseradish peroxidase was also a Sigma product,Type I, 20%. The mol wt of the enzyme was taken as40,000. Samples of kanamycin sulfate were obtained from H.Lundbeck & Co., Copenhagen, gentamicin sulfate fromSchering Diagnostics, Port Reading, N. J., and carbama-zepine from Ciba-Geigy AG, Basel. Polymyxin B sulfatewas from Novo Industry A/S, Copenhagen.

Bilirubin-albumin solution. 87 mg human serum albuminwas dissolved in 20 ml water and the pH adjusted to about9 with NaOH. Bilirubin, 4.0 mg, was dissolved in 200 ,ul0.5 M NaOH and 5 ml water. HCl, 75 ,uI, 1 M, wasadded, and 0.5 ml of this solution was immediately mixedwith the albumin. The mixture was kept in the dark andwas stable for 5 days in the refrigerator, giving no sig-nificant changes of light absorption spectrum or oxidationvelocity. Reproducible light absorption spectra and oxidationvelocities were obtained when albumin and bilirubin weremixed at a slightly alkaline pH, as above, avoiding colloidformation, which takes place if the bilirubin is added toa neutral albumin solution.

Exposure to bright daylight was avoided.Procedure. 1 ml bilirubin-albumin solution and 1 ml

sodium phosphate buffer, 133 mnM, pH 7.40, containing thedrug to be tested, was mixed. The final pH was 7.40±0.05.The solution was placed in a 1-cm cell in the BeckmanActa CV spectrophotometer (Beckman Instruments, Inc.,Fullerton, Calif.) with thermostatic cell holder at 37.0+0.2°C. A light absorption spectrum was recorded between580 and 350 nm. Absorbance at maximum (460 nm) wasclose to 0.81 in the absence of a binding drug, correspond-ing to a final bilirubin concentration of 15 MM in the re-action mixture.

The concentration of free bilirubin was about 0.015 MU,(17) increasing on addition of a strongly displacing drugbeyond the limit of solubility, which is about 0.1 MuM (24).In most cases, the pigment remained in a supersaturatedsolution. Colloid precipitation could be discovered by aprogressive decrease of maximal light absorption and broad-ening of the spectrum. In selected cases, the stability of thesolution was checked by measuring light scattering (23).

Hydrogen peroxide, 5 ,ul, 35 mM, freshly diluted, wasadded, and the light absorption was recorded as a functionof time at constant wavelength, 455 nm, during 3 min. Theabsorbance should remain constant.

Peroxidase solution, 100 4l in water, was added. A finalperoxidase concentration of 12 nM was used in experi-ments without displacement, and lower concentrations asneeded with added drug. The event marker was activatedat the moment of peroxidase addition, and recording of theabsorbance at 455 nm was continued until below 0.55.

Stability of the hydrogen peroxide was checked by addi-tion of another 5 Mul of hydrogen peroxide solution at thispoint. No change of reaction rate was observed. The finalabsorbance, about 0.012, was reached after several hours,or at 45-50°C in 1 h.

The shape of the spectral curves varied somewhat withseveral of the drugs. The wavelength used for recordingthe progress of the oxidation, 455 nm, in all cases was close

to the maximum. The final optical density after completedoxidation was not influenced much by the drugs.

The time taken for oxidation of the initial 0.2 of thetotal amount of bilirubin was determined from the graph.The peroxidase concentration was chosen so that to.2 was

longer than 2 min, to keep within the range where thevelocity is proportional to the enzyme concentration, andshorter than 10 min, for stability.

Control test without albumini. To examine the possibleinfluence of the drug on the activity of the enzyme, parallelexperiments were carried out without albumin. A bilirubinconcentration of 1.2 MM was chosen, since lower concen-trations were unstable and precipitation was common athigher levels. Also, the amount of bilirubin dimer is neg-ligible in the diluted solution (25). 1 ml water and 1 mlbuffer, with or without drug, was mixed with 2 Ml alkalinebilirubin solution, 1.29 mM. Peroxidase concentration was24 pM. Light absorption spectrum and time course wererecorded at the sensitivity range 0.1 optical absorbance. Anyuntoward reaction taking place between the reactants,such as oxidation of the drug with hydrogen peroxide andperoxidase or reactions between drug and bilirubin, wouldbe discovered in the test, in addition to interaction of thedrug with the enzyme.

This control test is more difficult and less accurate thanthe test proper with albumin present. A high-performancespectrophotometer is needed for recording the low opticaldensities. The small volume of bilirubin stock solution, lim-ited stability of bilirubin in the absence of albumin, anddeposition of the pigment on the walls of the cuvette tendto limit the reproducibility. Variations occurred in therange of 15% of the rate of oxidation. The lack of sensi-tivity of the control test could in several cases be counter-balanced by using higher concentration of the drug, thusassuring inactivity of the concentrations used in the testproper. In special cases, the inaccuracy of the control testmay theoretically be limiting for the determination of KD.This, however, is probably not of major practical impor-tance, since none of the 20 drugs tested here showed anyeffect in the control test.

Notes on the procedure. The hydrogen peroxide con-centration, calculated from the added amount, was initially80 MM. Determinations with an o-dianisidine reagent showedthat it remained above 25 MM. The velocity was constantfrom below 10 MMhydrogen peroxide and up to 200 MM.Higher concentrations were inhibitory.

Drugs of low solubility were added in alcoholic solutionto the buffered bilirubin-albumin solution. Ethanol, 1%vol/vol, influenced the oxidation rates only slightly. In criti-cal cases, tests were done with low concentrations of albu-min, 6 MM, and bilirubin, 3 MM, in a 4-cm cell, and per-oxidase concentration 2.4 nM or less (Fig. 4, phenylsalicylate). The low concentration of bilirubin did not giverise to instability when albumin was present.

The suitability of the enzymatic system and the albuminpreparation was tested by examining the variation of theoxidation velocity with albumin concentration (Fig. 2).Some albumin preparations, stored dry for about 1 yr at4°C, gave increased velocities at high albumin concentra-tions. Defatted albumin was more stable during storage thanalbumin with natural fatty acids.

The reproducibility of the oxidation velocity determina-tions without added drug corresponded to a standard devia-tion of about ±3%o, increasing to the order of ±5%c in thepresence of a strongly competing drug.

Competitive Binding of Bilirubin and Drugs to Human Serum Albumin 1305

0.5 /

_ \t~~~~~~~~~~450 500

FIGURE 1 The peroxidase method. Full linessorption spectrum (left) of a solution containhuman serum albumin, 15 juM bilirubin, and 50gen peroxide in phosphate buffer, pH 7.4. 1(right) of the maximal extinction, before, andtion of peroxidase at zero time. The thin hoiindicates the calculated extinction after dilutiorthe addition of enzyme solution. The stippled 4obtained after previous irradiation of the bilirtsolution for 3 min in sunlight behind 1-cm wind(

RESULTSInvestigation of the perozidase method.

of the oxidation of bilirubin, in a solution con'mol albumin/mol bilirubin, with hydrogen pe

50

[HRPI mn

-0 5x1FIGURE 2 The initial velocity, relative to the ecentration, of the oxidation of bilirubin wilWperoxide and peroxidase (ordinates), in the prmolar excess of albumin, as a function of the ztein excess (abscissa). pH 7.4, 37°C.

%0.2 min

I I HRP I M-1

4x107

W___.. FIGuRE 3 The time, to.,, taken for oxidation of 0.2 of the550 nm total amount of bilirubin (ordinates), as a function of the

light ab- reverse enzyme concentration (abscissa). pH 7.4, 370C.sing 30 iMAM hydro- peroxidase, is seen in Fig. 1 (insert, full line). The op-

rime course tical density at 455 nm was recorded during 5 min beforeL after addi- addition of the enzyme and remained constant, indicatingrizcntal liby stability of the system. Crude albumin preparations withcurves were a high content of hemoglobin gave a declining curvebin-albumin (16). On addition of the enzyme, the optical densityow glass. initially decreased (to the level of the thin line, calcu-

lated) due to the dilution, and then proceeded to de-cline with somewhat decreasing velocity. A final levelat optical density around 0.12 was reached in several

The course hours. Oxidation is seen to result in product(s) withtaining two less light absorption at this wavelength. The chemicalbroxide and nature of the process is not known (16). The shape of

the curve indicates that intermediate products with highoptical density are not formed or are short-lived. Oxida-tion of bilirubin with hydrogen peroxide and peroxidasein this respect seems to be different from photooxidation,which proceeds through strongly colored intermediates(26-28). Monitoring of the optical density at 455 nmapparently is suitable for observing the progress of theenzymatic oxidation.

Exposure of the bilirubin-albumin mixture to sunlightcaused a slight decrease of the light absorption spectrum(Fig. 1, left, stippled curve). When the irradiated mix-ture was oxidized with hydrogen peroxide and peroxi-dase, the course of the optical density was initially

M-1 steeper (insert, stippled curve). For practical purposes,P- 1 this shows that exposure to strong light should be

avoided. It seems possible that photodecomposition re-sults in accumulation of a product that absorbs light at455 nm and that is oxidized rapidly if hydrogen peroxide

enzyme con- and peroxidase are added. This product, if it occurs ash hydrogen an intermediate in the enzymatic process, is short-livedresence of a

and does not interfere with determination of the velocityof the rate-limiting step.

1356 R. Brodersen

\1%\\

In the absence of a competing drug, the velocity ofoxidation of bilirubin with hydrogen peroxide and perox-idase in mixtures with albumin decreases with increas-ing albumin concentration. This variation was examinedand the results are pictured in Fig. 2. The added concen-

tration of bilirubin was constant, 15 AM, and the concen-tration of albumin was varied from 30 to 120 AM. Freebilirubin concentrations in these solutions were of theorder of 0.01 1AM or less, since the binding constant is ofthe order 108 M-1. The concentration of free bilirubin was

FIGuRE 4 Free bilirubin concentrations, relative to the value without a displacer, measuredas tO.2(C)[HRPI /(tO.2[HRP]), as a function of displacing drug concentration (in millimolesper liter). The linear course is consistent with a simple, competitive 1: 1 displacement Theslope of the line is equal to the association constant of the drug at the high-affinity site forbilirubin. 0, Albumin 30 1AM, bilirubin 15 IAM, 1-cm cell. Ol, Albumin 6 1AM, bilirubin 3 IAM,4-cm cell. pH 7.4, 370C.

Competitive Binding of Bilirubin and Drugs to Human Serum Albumin 1357

TABLE IBinding Constants of Drugs to the High-Affinity

Bilirubin-Binding Site of HumanSerumAlubmin, at pH 7.4, 370C, with 60 mM

Sodium Phosphate Buffer

Drug

SulfisoxazoleSulfamethizoleSulfadiazineSulfanilamide

Sodium salicylateAcetylsalicylic acidPhenyl salicylateSodium benzoateMethyl p-hydroxybenzoatePropyl p-hydroxybenzoate

Benzylpenicillin sodiumKanamycin sulfateGentamicin sulfatePolymyxin B sulfate

DiphenylhydantoinCarbamazepinPhenobarbital sodium

PhenylbutazoneN-acetyl-L-tryptophanSodium orotate

KD

31-1

1.9 X 1048.6 X 1038.2 X 102<2 X 10'

9.2 X 10'5.3 X 1022.9 X 1041.7 X 1025.7 X 1045.3 X 104

1.5 X 10'<1 X 101<1 X 101<1 X 102

2 X 10'2.7 X 10'6.3 X 102

4.3 X 1041.1 X 103<2 X 102

1/[HRP]. This was fulfilled with a good approximationat moderate velocities, as used in the present method. Atvery high rates, a delay of about 4 s was found, althoughthis was of borderline significance. The line in Fig. 3therefore intersects the ordinate axis slightly above theorigin. It can easily be shown that this line is the ex-pected result if dissociation occurs as a first-order proc-ess, with a velocity constant of 3 min'. This shows thatthe dissociation is so fast that the oxidation can be re-garded as rate determining with good approximation, ifthe time taken for 0.2 of the total oxidation is kepthigher than 2 min.

At low enzyme concentrations (not shown in thefigure) the range of proportionality extends below 20pM peroxidase. The enzyme is remarkably stable evenat these extreme dilutions.

Competitive effect of drugs. The competitive effect of20 substances was examined. The method proved applica-ble in all cases. Two cationic drugs, gentamicin andpolymyxin B, precipitated bilirubin in the absence ofalbumin. These substances did not change the rate ofoxidation of bilirubin when albumin was present, andit was accordingly possible to ascertain their noncom-petitive nature. No other complications were encountered.

15 of the drugs showed measurable competition withbilirubin. In all cases a linear relationship was obtainedfor the velocity as a function of drug concentration (Fig.4), in agreement with the proposed model.

It is concluded that certain drugs decrease the bind-ing of bilirubin to its high-affinity site on human serum

very small compared to that of bound. In the equilibriumequation, the total bilirubin concentration could thus beused in place of the bound. This gives

B XKB P-B

(5)

If the velocity of the oxidation process is proportional tothe free bilirubin concentration b, and B and KB areconstant, we would expect to find a velocity proportionalto 1/(P-B). As seen in Fig. 2, this was experimentallyverified. This seems to confirm that a low concentrationof free bilirubin is actually present in equilibrium withthe bound. The oxidation apparently involved the freebilirubin only. Oxidation of the bound pigment wouldhave resulted in a line with a positive intercept at theordinate.

Variation of the oxidation velocity with the enzymeconcentration was examined, with constant concentra-tions of bilirubin and albumin (Fig. 3). If the dissocia-tion of bilirubin from the complex with albumin is veryfast, compared to the oxidation, so that the latter processis rate limiting, the velocity should be proportional tothe enzyme concentration, which means that to.2 shouldbe proportional to the inverse enzyme concentration,

FIGURE 5 Light absorption spectra of bilirubin-albumin,15/30 /AM, in phosphate buffer, pH 7.4, in the presence ofpolymyxin B, concentrations 0, 0.4, 0.7, and 1.0 mM. Cell,1 cm.

1358 R. Brodersen

albumin. The increase of free bilirubin concentrationmay be described in terms of competitive binding of thedrug to this site. The binding constants are seen inTable I.

Binding of drugs to other sites. Certain ligands maybind to albumin without interfering with bilirubin bind-ing. N-acetyl-L-tryptophan binds to one site on humanserum albumin at pH 7.4, with the association constant3 X 10' M-1 (29). The affinity for the bilirubin site islow, KD= 1.1 X 108 M' (Table I). This seems to indi-cate that the first molecule of acetyltryptophan is boundto a locus other than the first site for bilirubin.

Shifts of the spectrum of bilirubin-albumin were seenwith most of the drugs tested, even when the free bili-rubin concentration remained constant. An example ofthis is shown in Fig. 5. The spectrum of a bilirubin-albumin solution, 15 and 30 iM, is gradually changed byaddition of polymyxin B, 0.4-1 mM. The shoulder at420 nm is decreased and the maximum at 460 nm is in-creased. Some drugs (salicylate and related substances)caused a shift in the opposite direction. Polymyxin Band several other drugs apparently are bound to albu-min at sites other than the high-affinity site for bilirubin.It is interesting to note that binding of polymyxin B toone site influences the light absorption spectrum of bili-rubin, bound to a different site, without causing a changein the affinity of the bilirubin-albumin bond.

Besides the competitive binding of drugs demonstratedabove, another type of liganding of some drugs takesplace at sites distinct from the first site for bilirubin.This does not seem to influence the affinity of bilirubinfor the first site. A third, hypothetical, type of liganding,in which a drug bound to one site decreases or in-creases the affinity of bilirubin on another site, wouldgive curved or sigmoid-shaped relationships and hasnot been demonstrated in the present material.

DISCUSSIONThe experimental principle. Evaluation of kernic-

terus-promoting effects of drugs, as caused by inter-action with binding of bilirubin to albumin, cannot bedone directly on newborn humans, and the use of ex-perimental animals is complicated by species variationof binding patterns for drugs and bilirubin to serum al-bumin. In vitro studies of binding interactions must berelied upon and the effects in the fetus or newborn childhas to be inferred from the in vitro observations. Forthis purpose, knowledge of the mechanism of interaction,whether reversible, competitive, etc., is essential. Fur-thermore, a quantitative description of the interaction isnecessary.

A drug not bound to albumin cannot function as acompetitor. Binding is a necessary but not a sufficientcondition for a bilirubin-displacing effect, since bili-

rubin and drug may ligand to distinct sites on the al-bumin molecule. This seems in fact to take place withsome drugs. It is consequently necessary to study theshifts of binding equilibrium caused by the drug, and thisagain means that knowledge of the physical state ofbilirubin in solution is prerequisite.

At a pH of about 8-9, bilirubin is soluble in waterand is present as a monomer anion in equilibrium withits dimer (25). The dissociation constant is of theorder of 10-' M, independent of pH, indicating thatbilirubin in the micro- or nanomolar range is predomi-nantly monomer. At pH 7.4, the solubility is very lowand colloid aggregation tends to take place at concen-trations higher than about 0.1 /M (24), although higherconcentrations may remain in a metastable, supersatu-rated solution for some time. In the presence of serumalbumin, reversible binding of bilirubin takes place toone high-affinity and a few weaker sites (17, 23). Ac-cording to general concepts of binding equilibrium, thisshould result in a certain concentration of unbound bili-rubin in the solution. At a ratio 1: 2 of bilirubin to al-bumin, the free bilirubin concentration is very low, prob-ably of the order of 10 nM (17), and in any case verysmall compared to the concentration of bound pigment.Under these circumstances it is not possible to measurethe free bilirubin concentration by conventional, spec-troscopic means, and evidence for its existence can onlybe obtained indirectly. The reality of such small, free,monomer bilirubin concentrations is supported by thefindings in the present paper of predicted oxidationvelocity relations to the concentration of albumin (Fig.2) and competing drugs (Fig. 4).

Various experimental principles have been used fordemonstration of the bilirubin-displacing effect of drugs.A shift of light absorption spectrum takes place whenbilirubin is bound to albumin. This, however, cannot beused as a basis for studying binding equilibria at thefirst site, due to the very low ratio of free to bound pig-ment. For the same reason, it is also not possible to uti-lize the fact that bound bilirubin shows a fluorescencewhile the free does not. The same is true for the opti-cal rotation, seen with bound bilirubin, but not with thefree. All these principles would be further complicatedby shifts of the spectra that occur on binding of newligands, drugs, to other sites on the bilirubin-albumincomplex (Fig. 5).

Separation of the free bilirubin by ultrafiltration, equi-librium dialysis, or ultracentrifugation (30), followed byspectrophotometric determination, gives disappointing re-stlts due to instability of bilirubin in very dilute solutionsand a tendency of the pigment to adhere to membranesand glass surfaces. Attempts were made in the author'slaboratory to use 'C-labeled bilirubin in an ultracentri-fugation technique. It was found that a small, nearly

Competitive Binding of Bilirubin and Drugs to Human Serum Albumin 1359

constant, fraction of the radioactivity was present inthe protein-free supernate. This could not be identifiedas bilirubin and was thought to originate from decom-position, probably due to light (26).

Observation of bilirubin displacement to an adsorbent,cholestyramine (9, 28) or Sephadex (12, 19, 31, 32), orto erythrocytes (7, 8, 33, 34) has yielded valuable in-formation. Displacement by several drugs has been dem-onstrated, but these methods have not been used forquantitative determination of the binding constants.

The peroxidase method is based on an analogous prin-ciple. Two proteins, peroxidase and albumin, compete forthe bilirubin. Added drugs displace the ligand from albu-min to the enzyme, resulting in an increased rate of oxida-tion. Several advantages are obtained by use of peroxi-dase as the competing binder, rather than an adsorbentlike Sephadex. First, the fraction of the total bilirubinbound to the enzyme is extremely small so that the equi-librium between bilirubin, drug, and albumin, is undis-turbed (in the initial moment of the oxidation process).Second, the amount of bilirubin bound to peroxidase isproportional to the free bilirubin concentration and iseasily quantified by measuring the initial rate of oxida-tion. Finally, it is possible to examine whether the drugcompetes with bilirubin for binding to the enzyme bytesting the influence of the drug on the oxidation velocityof bilirubin in the absence of albumin.

The mechanism of displacement of bilirubin with thedrugs tested here seems to be sufficiently described bythe model of competitive binding. The experimental re-sults of the present work may be interpreted in termsof simple competition of one molecule of drug with oneof bilirubin for the high-affinity bilirubin site on thealbumin molecule. Other stoichiometric relations, suchas displacement of one molecule of bilirubin by two mole-cules of the drug, would give curved lines for the varia-tion of oxidation velocity with drug concentration.

Site-to-site inhibition or cooperativity of binding, ef-fected through conformational changes, would likewiseresult in curved functions. Such effects have been ob-served when long-chain fatty acids (35), steroids (36),detergents (37), or calcium ions (36) are bound toserum albumin. It is, accordingly, necessary with eachdisplacing substance to examine whether a linear rela-tionship of oxidation velocity and drug concentrationexists at levels of the latter considerably higher than thatof albumin, to establish a simple, competitive mechanismof displacement.

Description of the mechanism of displacement as asimple competition between one molecule of bilirubinand one of the drug for a specific site on the albuminmolecule opens the possibility of assessing the effect ofeach drug by its binding constant, KD, to this locus.Knowledge of this constant and of the free drug concen-

tration in plasma defines the relative increase of freebilirubin concentration caused by the drug, according toEq. 1. It is not necessary to know the affinity of bili-rubin for binding to albumin.

Validity of the approximative assumptions. The ex-perimental determination of KD by the peroxidase methodis based on several approximations, namely: (a) Thefree bilirubin concentration is very small compared tothe total; ( b) The free concentration of the competingdrug is equal to the total; and (c) Bilirubin is boundto one site only.

The errors caused hereby in determination of KD arediscussed in the following:(a) In a solution containing 30 WMalbumin and 15 uM

bilirubin, the free concentration of the latter is equalto 1/KB, i.e., probably around 10 M. With a strongcompetitor at high concentration, it might appear pos-sible to displace a major fraction of the bilirubin. Thiswould result in smaller increases of oxidation velocityat high drug concentrations. The straight lines obtainedin all cases, with three of the drugs up to free bilirubinconcentrations as high as 50 times the original (Fig.4, methyl p-hydroxygenzoate, propyl p-hydroxybenzoate,phenylbutazone), confirm the validity of the assumptionthat the free bilirubin concentration is very small com-pared to the bound within this range of increase. (b)Eq. 1 is thus valid in all cases and may be written:

b = KDdD+ 1.bo D (6)

At high drug levels, the free concentration, d, approachesthe total, D, Eq. 2 becomes valid, and the slope of theline approaches KD. At the other extreme, when D issmall compared to the albumin concentration, the slopeis smaller, KDd/D.

Let us first consider the deviation if the drug is boundstrongly to one or a few sites. A curve is then obtained(Fig. 6a), merging into a straight line with the slopeKD at high drug concentrations. This straight line isshifted along the abscissa away from the ideal, by a cer-tain concentration, e. The value of e is equal to the maxi-mal concentration, nP - B., of bound drug, obtained atfull saturation of all sites, where n is the total numberof drug-binding sites, since the concentration, B, of theseis occupied by bilirubin. A straight line, fitted through(0, 1) and the experimental points scattered along thecurve, may be drawn through a point on the straightportion of the curve around the abscissa Dm../2, if Dina.is the highest concentration of the drug plotted in thegraph. (Fig. 6a, dotted line). The slope of this line isnot KD, but

1_2(nP- B))KDna l

1360 R. Brodersen

This makes it possible to estimate the error of the de-termination of KD. The greatest error is obtained with ahigh value of n and low Dm.&, relative to the albuminconcentration. Among the 20 drugs tested (Fig. 4),diphenylhydantoin would be likely to show an especiallylarge error. At a guessed n value of 4, KD would bedetermined as low as 0.475 times the true value. It isunlikely, however, that a corresponding curvature ofthe line should be masked by experimental scatter of thepoints. Actual errors are probably smaller. In case ofdoubt, the determination could be repeated with one-fifth the concentrations of albumin and bilirubin, re-ducing the error to one fifth. This was done in case ofdiphenylhydantoin, but gave no significant change ofthe slope, (Fig. 7). Typical errors in this categoryrange about 1-10%.

If the drug is bound with low affinity to a high num-ber of loci, in addition to the bilirubin site, it becomesimpossible to attain drug concentrations high enough tofulfill the condition d D. This would be likely to re-sult in a curved line, except in the extreme case whenall sites remain largely unoccupied but are present in ahigh number, so that the total amount of bound drugbecomes significant. The ratio of free to bound drugwould then remain constant at all concentrations, and astraight line would ensue with the slope KDd/D, result-ing in a corresponding error of KD. A 50% error wouldbe caused by binding to 170 sites, each occupied to onetenth, if the highest drug concentration is 1 mM, andwith 30 iM albumin. With 6 /AM albumin, 850 siteswould be needed. Errors of the order of 1-10% are morelikely.

(c) Secondary binding sites for bilirubin are presenton the albumin molecule (17, 23), and may influence thedetermination of KD. The effect of a competing drugunder the circumstances of the in vitro experiment maybe described as a change of the binding constant forbilirubin to a lower, apparent value,

KB(app) = KB!(KDd + 1), (7)

derived in analogy with the apparent Michaelis constantof an enzyme-substrate in the presence of a competitiveinhibitor. Corresponding relations hold for the secondarysites. If KD values of all sites are equal, it is seen thatthe free bilirubin concentration will vary with that ofthe drug in accordance with Eq. 1. The true value ofKD is then obtained from the slope of the line. Problemsarise when the KD of the primary site is much higherthan that of the secondary, i.e. when the drug com-petes mainly or exclusively for the primary site. Inthis case, the apparent KB of the primary site mayapproach the values of the secondary loci, or the ap-parent affinities may even be reversed, with the resultthat major amounts of bilirubin are displaced to the

FIGuRE 6 Deviations from the straight linear course ofthe relation of free bilirubin to added drug concentration.In a, a curvature has resulted from binding of the drugto one or several strong sites on the albumin molecule. bshows the type of deviation expected if substantial amountsof bilirubin are displaced by the drug to weaker sites.

secondary sites. Each site will then carry a smalleramount of bilirubin, and the free bilirubin concentra-tion will be lower than predicted from Eq. 1. At lowdrug concentrations, the slope of the line will be closeto KD for the primary site, provided that nearly allbilirubin is bound to this in the absence of a drug. Atvery high drug levels, the slope will approach the low-est KD value of any site, since all bilirubin finally willbe displaced to the site where the drug has the leastaffinity (Fig. 6b). The slope may theoretically ap-proach zero if one or more secondary sites do not bindthe drug at all. Curvatures of this type are best ob-served if the examinations are extended to high drugconcentrations, as permitted by solubility, etc. Duringthe work with the limited number of drugs in thepresent paper, no such observations were made. The

2

1.5

1

-b

Qo/0O

0

0.1 0.2 0.3 0.4!~:b I

0 mM

FIGURE 7 Detail of Fig. 4: diphenylhydantoin. Resultsobtained with 6 and with 30 ,uM albumin are identical

Competitive Binding of Bilirubin and Drugs to Human Serum Albumin 1361

TABLE IICalculated Total and Free Bilirubin Concentration in Blood Plasma in the Presence of a Competing Drug,

and the Resultant Displacement of Bilirubin from Albumin to a Tissue Receptor

Bilirubin

After equilibration with tissuesDrugFree Concn. in plasma Amount

concn. (drug added in vitro) Concn. in plasma bound toin nonalbumin Amount

plasma Bound Free Bound Free receptor displaced

mM mg/100 ml pM pM pM FM p0mol pmol

0 17.5200 300.000 0.043 300.0 0.043 0.1 00.01 17.5197 299.996 0.047 299.9 0.047 0.11 0.010.1 17.5174 299.957 0.086 299 0.086 0.2 0.11 17.4923 299.527 0.47 290 0.45 1.1 1

10 17.27 295.7 4.3 ca 200 ca 4 ca 10 ca 10

Total plasma albumin concn. 400 AM, plasma volume 100 ml, KB = 7 X 107 M-1, KD = 1.0 X 104 M-1.Displacement to or from reservoirs other than plasma albumin and an arbitrarily chosen receptor isneglected.

reason for this may be that the affinity of bilirubinfor the secondary sites is very much less than for thefirst. It is possible, however, that curvatures of thetypes shown in Fig. 6a and b may partly cancel eachother under suitable circumstances. If, therefore, a drugshows a degree of competition, estimated to be of clini-cal importance (see below), further studies of bindingto all sites on albumin should be conducted to evaluatethe possible errors. It should be noticed that all sourcesof error tend to give too low KD values.

Attempted calculation of maximal displacement invivo. The binding constants for drugs to the high-affinity bilirubin site fall in the same range as bindingconstants for most drugs to albumin, 0-10w M-1, asknown from studies of equilibrium dialysis, etc. (38),and are several orders of magnitude lower than thatof bilirubin, about 10' M'. Generally the affinities areunfavorable for displacement of bilirubin with drugs.Accordingly, addition of a drug to blood plasma in vitrodoes not result in release of a major fraction of thebound bilirubin. The relative increase of the free bili-rubin concentration may, however, be high. This isillustrated in Table II, by calculated values of boundand free bilirubin concentrations in plasma with vary-ing concentrations of a competing drug. At zero drugconcentration, 300 FM albumin of the 400 present isoccupied by bilirubin, leaving 100 MM free albumin.The free bilirubin concentration is calculated from theequilibrium equation, 300 X 10V/(b x 100 X 10-) = 7X 107. With added drug, the free bilirubin concentra-tions are calculated from Eq. 1.

From the left-hand half of Table II, it is obviousthat the effect of the drug in vitro is not a gross dis-placement of bilirubin from binding to albumin but

consists of occupation of available sites, whereby thefree bilirubin concentration is increased in the sameproportion as the amount of vacant sites is decreased.The concentration of bound or total bilirubin is notdecreased measurably in the absence of nonalbuminreceptors that can take over the bilirubin.

Displacement of a large fraction of the bilirubinmay result in vivo. According to Odell (39), freebilirubin in the blood plasma is diffusible and inequilibrium with receptors outside the circulation. Anincrease of the free bilirubin concentration in theplasma results in an increased extravascular binding.Numerical evaluation of the displacement is possiblein certain conditions, as follows:

Let P* denote the molar concentration of a tissueacceptor site and d* etc. signify the other parametersof the tissue phase, in analogy with the symbols usedfor blood plasma. The free bilirubin concentration inthe plasma may be obtained from the former equationsas

b =I Xp [BP] (KDd+ 1).KB P -[BP] D

In equilibrium, the same free bilirubin concentrationprevails in the tissue, and this is found from the analo-gous expression

b = *X [BP* (KD*d* + 1)KB* P* - [BP*]

and in the absence of a drug,

b I= X [BP*]oKB* P* - [BP*].o

1362 R. Brodersen

From the latter two equations we derive,

r[BP*] b KB*bo + 1[BP*]o bo KD*d* + KB*b + 1

which is equal to the relative increase of bilirubin,bound to the acceptor, and is a measure of the displace-ment. It is seen that the maximal value of this is ob-tained if KD* << 1/d* and b << 1/KB*, i.e., if thedrug does not displace bilirubin from the tissue ac-ceptor, and if the acceptor is far from saturated withbilirubin. In this case, we have, using Eq. 1, and bo < b,

(EBP*] = KDd + 1.[BP*]o / max

This maximal displacement factor is a quantitativemeasure of the risk of bilirubin displacement incurredby the use of a drug at a certain free concentrationin the plasma. It may be convenient to signify this fig-ure by 5 and to express it by the total plasma concen-tration D of the drug,

6 = KDd + 1 = KD(1-q)D + 1, (8)

when q is the fraction of the drug bound to plasmaproteins.

The numerical value may be calculated if KD isknown and the free drug concentration needed fortherapeutic effect or the total plasma concentrationand degree of protein binding are specified.

The binding constants obtained for the four sulfon-amides (Table I) indicate an increasing competitionwith bilirubin in the order sulfanilamide < sulfadiazine< sulfamethizole < sulfisoxazole. Bratlid (7) has ex-amined the binding of bilirubin to erythrocytes in solu-tions with albumin after addition of these sulfonamidesand has found that it is increased in the same order ofefficiency. Sulfisoxazole at usual dosage gives a plasmalevel about 0.5 mM, 85% of which is protein bound.The maximal displacement factor, calculated from Eq.8, is 2.4. At this blood plasma concentration, sulfisoxa-zole may thus cause an increase by a factor of 2.4 ofthe amount of bilirubin bound to a critical receptor,such as an area of the brain. These findings are con-sistent with the clinical experience of a displacingeffect of sulfisoxazole (1, 2, 5).

The mechanism of displacement with salicylate. So-dium salicylate is a fairly potent competitor, as seenfrom its KD value, 0.9 X 10' M-1. The degree of pro-tein binding, however, is high at moderate plasma con-centrations. According to Rudman, Bixler, and DelRio (35), two molecules of salicylate are bound tohuman serum albumin with a binding constant of 105M'. Half-saturation of these sites is produced bymoderate doses and gives a free plasma concentrationof 10-' M of salicylate. The maximal displacement fac-

tor is 1.09, in agreement with the fact that clinical dis-placement has not been reported with this drug. Pre-sumably, the first two molecules of salicylate are boundto human serum albumin without interfering with thebilirubin site. A following molecule of the drug, boundin a subordinate class of sites, competitively displacesbilirubin. Maximal dosage, giving a drug/albumin ratioof 3 or 4, would be expected to result in displacement.A decrease of bilirubin concentration in blood plasmahas in fact been demonstrated after large doses of salic-ylate in the Gunn rat (9, 10) and in man (8).

Thaler and Schmid (14) refer to the permanentacetylation of albumin with acetylsalicylic acid, dem-onstrated by Hawkins, Pinckard, Crawford, and Farr(15), and warn against use of this drug in patientswith unconjugated hyperbilirubinemia. This is basedon the fact that acetylation of albumin decreases theaffinity for certain anions. Jacobsen (40) has found,however, that acetylation of human serum albumin invitro with acetylsalicylic acid did not change the af-finity for bilirubin. One specific lysyl-amino group isattacked by acetylsalicylic acid and seems to be in-volved in the binding of fatty acids and tryptophan(41), whereas bilirubin is bound to a different site.The competitive effect of acetylsalicylic acid as suchis slight (Table I), and large doses would be neededto produce displacing amounts of salicylate.

Clinical experience and observations from animal ex-periments and from kinetic studies in vitro can be ex-plained by a simple, competitive mechanism of bili-rubin displacement by drugs.

As stressed by Stern (42), any drug used in thenewborn should be examined for its ability to displacebilirubin from albumin. Even during pregnancy, a dis-placing drug may reach the fetus via the placenta andexert its effect in case of fetal hyperbilirubinemia.Similarly, passage of pharmaca with breast milk shouldbe considered. At present, such testing is not requiredby official authorities but ought to be carried out for alldrugs in current use and as a routine before licensingof new substances.

ACKNOWLEDGMENTSThe skillful technical work of Signe Andersen is acknowl-edged.

REFERENCES1. Silverman, W. A., D. H. Andersen, W. A. Blanc, and

D. N. Crozier. 1956. A difference in mortality rate andincidence of kernicterus among premature infants al-lotted to two prophylactic antibacterial regimens. Pedi-atrics. 18: 614-625.

2. Harris, R. C., J. F. Lucey. and J. R. MacLean. 1958.Kernicterus in premature infants associated with lowconcentrations of bilirubin in the plasma. Pediatrics.21: 875-884.

3. Johnson, L., F. Sarmiento, W. A. Blanc, and R. Day.

Competitive Binding of Bilirubin and Drugs to Human Serum Albumin 1363

1959. Kernicterus in rats with an inherited deficiencyof glucuronyl transferase. Am. J. Dis. Child. 97: 591-608.

4. Schutta, H. S., and L. Johnson. 1971. Electron micro-scopic observations on acute bilirubin encephalopathyin Gunn rats induced by sulphadimethoxine. Lab. Invest.24: 82-89.

5. Kantor, H. I., D. A. Sutherland, J. T. Leonard, J. H.Kamholz, N. D. Fry, and W. L. White. 1961. Effecton bilirubin metabolism in the newborn of sulfisoxazoleadministered to the mother. Obstet. Gynecol. 17: 494-500.

6. Odell, G. B. 1959. Studies in kernicterus. I. The pro-tein binding of bilirubin. J. Clin. Invest. 38: 823-833.

7. Bratlid, D. 1972. The effect of antimicrobial agents onbilirubin binding by human erythrocytes. Scand. J. Clin.Lab. Invest. 30: 331-337.

8. Watson, D. 1962. The transport of bile pigments: thebinding of sodium bilirubinate to plasma protein. Clin.Sci. (Oxf.). 22: 435 444.

9. Schmid, R., I. Diamond, L. Hammaker, and C. B.Gundersen. 1965. Interaction of bilirubin with albumin.Nature (Lond.). 206: 1041-1043.

10. Diamond, I., and R. Schmid. 1966. Experimental bili-rubin encephalopathy. The mode of entry of bilirubin-IC into the central nervous system. J. Clin. Invest. 45:678-689.

11. Zamet, P., and F. Chunga. 1971. Separation by gelfiltration and microdetermination of unbound bilirubin.II. Study of sera in icteric newborn infants. Acta Pae-diatr. Scand. 60: 33-38.

12. Josephson, B., and P. Furst. 1966. Sulphonamides com-peting with bilirubin for conjugation to albumin. Scand.J. Clin. Lab. Invest. 18: 51-63.

13. Wennberg, R. P. 1971. Albumin binding of bilirubin:Theoretical, experimental and clinical considerations.Proceedings of the Thirteenth International Congress onPediatrics. Wiener Medicinischen Akademie, Vienna,Austria. Vol. I: 311-316. (Abstr.).

14. Thaler, M., and R. Schmid. 1971. Drugs and bilirubin.Pediatrics. 47: 807-810.

15. Hawkins, D., R N. Pinckard, I. P. Crawford, and R. S.Farr. 1969. Structural changes in human serum albumininduced by ingestion of acetylsalicylic acid. J. Clin.Invest. 48: 536-542.

16. Brodersen, R., and P. Bartels. 1969. Enzymatic oxida-tion of bilirubin. Eur. J. Biochem. 10: 468-473.

17. Jacobsen, J. 1969. Binding of bilirubin to human serumalbumin. Determination of the dissociation constants.FEBS (Fed. Eur. Biochem. Soc.) Lett. 5: 112-114.

18. Dole, V. P. 1956. A relation between non-esterifiedfatty acids in plasma and the metabolism of glucose.J. Clin. Invest. 35: 150-154.

19. Chan, G., D. Schiff, and L. Stern. 1971. Competitivebinding of free fatty acids and bilirubin to albumin:Differences in HBABAdye versus Sephadex G-25 in-terpretation of results. Clin. Biochem. 4: 208-214.

20. Starinsky, B., and E. Shafrir. 1970. Displacement ofalbumin-bound bilirubin by free fatty acids. Implicationsfor neonatal hyperbilirubinemia. Clin. Chim. Acta. 29:311-318.

21. Thiessen, H., J. Jacobsen, and R. Brodersen. 1972. Dis-placement of albumin-bound bilirubin by fatty acids.Acta Paediatr. Scand. 61: 285-288.

22. Chen, R. F. 1967. Removal of fatty acids from serumalbumin by charcoal treatment. J. Biol. Chem. 242:173-181.

23. Brodersen, R., L. Funding, A. 0. Pedersen, and H.Rjigaard-Pedersen. 1972. Binding of bilirubin to low-affinity sites of human serum albumin in vitro fol-lowed by cocrystallization. Scand. J. Clin. Lab. Invest.29: 433-446.

24. Brodersen, R., and J. Theilgaard. 1969. Bilirubin col-loid formation in neutral aqueous solution. Scand. J.Clin. Lab. Invest. 24: 395-398.

25. Brodersen, R. 1966. Dimerisation of bilirubin anion inaqueous solution. Acta. Chem. Scand. 20: 2895-2896.

26. Kapitulnik, J., S. H. Blondheim, A. Grunfeld, and N. A.Kaufmann. 1973. Photodecomposition of bilirubin: ultra-filtrable derivatives. Clin. Chim. Acta. 47: 159-166.

27. Garbagnati, E., and P. Manitto. 1973. A new class ofbilirubin photoderivatives obtained in vitro and theirpossible formation in jaundiced infants. J. Pediatr. 83:109-115.

28. Odell, G. B., R. S. Brown, A. B. and N. A. Holtzman.1970. Dye-sensitized photooxidation of albumin asso-ciated with a decreased capacity for protein-binding ofbilirubin. Birth Defects. Orig. Artic. Ser. 6: 31-36.

29. McMenamy, R. H., and J. L. Oncley. 1958. The spe-cific binding of L-tryptophan to serum albumin. J. Biol.Chem. 233: 1436-1447.

30. Clarenburg, R., and J. L. Barnhart. 1973. Interactionof serum albumin and bilirubin at low concentrations.Am. J. Physiol. 225: 493-496.

31. Schiff, D., G. Chan, and L. Stern. 1971. Fixed drugcombinations and the displacement of bilirubin fromalbumin. Pediatrics. 48: 139-141.

32. Windorfer, A., Jr., and K. Mihailova. 1973. Der Ein-fluss verschiedener Medikamente und pH-Werte aufdie Albumin-Bilirubin-Bindung in vitro. Z. Kinderheilkd.114: 83-92.

33. Bratlid, D. 1973. Reserve albumin binding capacity,salicylate saturation index and red cell binding of bili-rubin in neonatal jaundice. Arch. Dis. Child. 48: 393-397.

34. Bratlid, D., and A. Langslet. 1973. Displacement of al-bumin-bound bilirubin by injectable diazepam prepara-tions in vitro. Acta Paediatr. Scand. 62: 510-512.

35. Rudman, D., T. J. Bixler, and A. E. Del Rio. 1971.Effect of free fatty acids on binding of drugs by bovineserum albumin, by human serum albumin and by rabbitserum. J. Pharmacol. Ezp. Ther. 176: 261-272.

36. Ryan, M. T. 1973. Influence of steroid binding on thetryptic hydrolysis of serum albumin. Biochemistry. 12:2221-2230.

37. Reynolds, J. A., J. P. Gallagher, and J. Steinhardt.1970. Effect of pH on the binding of N-alkyl sulfatesto bovine serum albumin. Biochemistry. 9: 1232-1238.

38. Meyer, M. C., and D. E. Guttman. 1968. The bindingof drugs by plasma proteins. J. Pharm. Sci. 57: 895-918.

39. Odell, G. B. 1959. The dissociation of bilirubin fromalbumin and its clinical implications. J. Pediatr. 55:268-279.

40. Jacobsen, C. 1972. Chemical modification of the high-affinity bilirubin-binding site of human-serum albumin.Eur. J. Biochem. 27: 513-519.

41. King, T. P., and M. Spencer. 1970. Structural studiesand organic ligand-binding properties of bovine plasmaalbumin. J. Biol. Chem. 245: 6134-6148.

42. Stern, L. 1972. Drug interactions. II. Drugs, the new-born infant, and the binding of bilirubin to albumin.Pediatrics. 49: 916-918.

1364 R. Brodersen